"Free" trade was among the top issues to dominate the U.S. presidential election, and it continues to be a sore point today. President Trump has excoriated such treaties as the North American Free Trade Agreement, and the now-rejected Trans-Pacific Partnership, under the assumption that they are the major reason for lost American manufacturing jobs. Others say that isn't so. In this conversation, excerpted from an episode of The SupplyChainBrain Podcast, we hear from Michael Hicks, professor of economics and director of the Center for Business and Economic Research at Ball State University. He explains why increased productivity has been a bigger contributor than trade to the fall in manufacturing employment, and where the real losses and gains have taken place.
Q: According to President Trump's website, citing research from the Economic Policy Institute, the U.S. has lost nearly a third of its manufacturing jobs since the North American Free Trade Agreement came into effect in 1994, and 50,000 factories since China joined the World Trade Organization. Other statistics show that 5.6 million manufacturing jobs have been lost in the U.S. between 2000 and 2010. Yet you don’t believe that this is primarily the result of trade or offshoring of manufacturing. What’s your perception regarding the reason behind those job losses?
Hicks: All of those data are correct. We’ve lost 5.5 million jobs since 2000 in manufacturing, about 7.2 million since peak annual manufacturing employment in 1979, and trade with China and Mexico has risen since Nafta took effect. What’s at question is the [impact] of Nafta, the WTO and trade-related losses since 2000, when China received permanent normalized trading relations with the U.S. Our studies and a number of others call into question whether or not most of those jobs have been lost to trade, or there’s something else going on — most particularly, the growth of productivity in manufacturing.
Q: In fact you state that 85 percent of those job losses can be attributed not to trade, but to automation and technological change. Is that accurate?
Hicks: Our study, which looked at manufacturing employment between 2000 and 2010, concluded that roughly 88 percent of job losses, give or take 5 percent either way, were due to increased productivity of manufacturing workers over that time period. Other studies have come up with similar numbers, looking at 1999-2013, from the peak of the pre-recession period to the manufacturing recovery that had occurred by 2013. The other study was by Daron Acemoglu of M.I.T. The two of us have been saying that 80 or 90 percent was due to productivity gains, and 15 to 20 percent due to trade.
Q: There was another M.I.T. study that found that rising Chinese imports between 1999 and 2011 cost up to 2.4 million American jobs. Is that in conflict with what you’re saying, or does that fit into the picture as you see it?
Hicks: There were two studies at M.I.T., one with Acemoglu, David Autor, Gordon Hanson and David Dorn, which [put losses at] about 2.4 million jobs, although actual manufacturing employment was closer to a million jobs during that time. These studies vary a little bit, based on the time period that you choose. It’s very difficult, because you go from the bottom of one recession to the bottom of another, you get one set of numbers. If you take a peak period to the bottom of a recession, you get different ones, because manufacturing tends to be very heavily affected by the business downturn. So the timing matters a bit. But these two studies have said that something like 10 to 20 percent of manufacturing jobs were lost due to trade, with the remainder due to automation or other productivity-related gains.
There are also studies that say that the data on productivity is misstated — maybe it’s 25 percent of manufacturing jobs lost due to trade. Others suggest that trade pressure on manufacturers may have accelerated the purchase of automated equipment, technology and leaner factories. In that way, trade had an indirect effect. But there’s nobody out there, other than a think tank like the Economic Progress Institute — which supported Bernie Sanders [for president], by the way — who says that anything more than about half the job losses were due to trade. It ranges from a low of single-digit percentages by some older studies all the way up to half, depending on how you indict the role of trade, either directly through import substitution, or indirectly by pressuring American businesses to become leaner over the past decade.
Q: Your own study is two years old. Has anything happened since then to change the validity of its conclusions?
Hicks: Our study was published in the summer of 2015, using data that ended around 2013. Between 2013 and the end of 2015, manufacturing employment continued to rise. Manufacturing production in the U.S. peaked in 2015 — we have good data on gross domestic product adjusted for inflation. So the point that American manufacturing workers made more in 2015 than in any previous year in history is certainly true. We don’t have data yet for 2016 on manufacturing GDP, but with little hints like automobile manufacturing peaking in December, we have some evidence that things were better.
Basically, two dynamics continue to occur: trade continues to matter an awful lot, with Americans buying more manufactured goods, but spending a smaller share of their income on them. And secondly, automation, improved workplace productivity and the education [level] of workers are all rising. Those complex factors tend to work together to mean that we’re going to make more manufactured goods in the U.S. We’re going to import and export more as well, but we’re probably not going to see employment growth in any significant way in the coming decades.
Q: How do you know in your study that automation has played such a big role? What kind of questions did you ask, and answers did you get?
Hicks: The Census Bureau does an annual census of manufacturing. There are very distinct data that are available and preserved in the joint scientific project [dating] back to 1958 with the National Bureau of Economic Research. There are data from the Bureau of Labor Statistics on employment, and secondary data that are reported through the Bureau of Economic Analysis. The numbers don’t match perfectly, but the trends are identical. We use the aggregate national data at each of the industry levels, then subtract out the growth of non-value-added product. Almost nothing is made in any one place — the phone I’m talking to you on was probably manufactured in five or six different countries, and assembled in the U.S. We have to net out the contribution of intermediate goods. We just track the growth in output per worker, and if you think about this, it shouldn’t be too surprising that what took, say, a hundred manufacturing workers in the 1970s to make is now being done with 40 or 50, in terms of inflation-adjusted value of goods produced. If we didn’t have Americans buying more goods, and foreigners buying more of our goods, we’d have a lot fewer manufacturing jobs than we now have.
Q: I’ve heard that in the so-called Second Industrial Revolution, the nation experienced a huge increase in worker productivity, but that more recently, productivity has not improved by anything close to that amount.
Hicks: It’s had very short periods of rapid growth. When economists talk about automation or productivity, we have in our minds some aggregate effect on the workplace. On the shop floor, it looks very different from this. Yes, there has been a stalling of manufacturing productivity growth — it’s still growing, but not at the 2.5 to 3.5-percent rate that occurred from the end of the Second World War through to about 2000. What the Bureau of Labor Statistics attempts to do is measure the quality of manufactured product, adjusted by what we buy. I’m talking to you on a phone with a handset that was installed at Ball State University in the mid-1990s, and probably cost a couple hundred dollars. Compare that piece of equipment to my cellphone, which was cheaper, and you can see how radically difficult it is to make good quality adjustments and value-added measures of a particular product. There’s no doubt that in the last four or five years, we’ve seen a productivity downswing, not just in manufacturing but more heavily in labor-intensive industries as well. That’s one of the puzzles of all this.
Q: Do you get a sense of what form this automation is taking, and which industries are most affected by it?
Hicks: It’s taken several forms. If you’re in steel production, it is a revolutionary shift away from the late 19th-century/early 20th century steel mills that were in Gary, Ind., Pittsburgh, Pa., and Birmingham, Ala. They’ve been replaced by mini-mills, so we’re actually producing as much American steel as we did in, say, 1960, but we’re doing it with about a quarter of the workforce. There’s been an obvious technological adjustment in how production works. If you’re on a shop floor of a factory and assembling something, it’s a little bit different. Part of it is robotics — automated equipment that is digitally managed (which I think, by the way, is the big future of all this). A lot of it is also about how the shop is set up. Has the union been able to featherbed jobs like they did in 1960? The answer is no, so maybe there’s 10 percent fewer workers who are sitting around idly, recovering from a hangover.
At the same time, in the warehouse, instead of having your workstation restocked by a lot of people who know where products are, maybe it’s being restocked by conveyor belt, an automated trolley or small truck with artificial intelligence that moves in and out of the warehouse. So that’s very different. Then if you go to a big processing plant, like for food or a warehouse at Amazon.com, there may be no people in the back. The tomatoes are unloaded, then sorted automatically by size. In the whole process, the only human intervention is somebody oiling the machines, fixing a stoppage or providing calibration if they think the temperature is off. The process can be both very automated and involving digital technology. It can affect the way labor works, and the number of people on the shop floor. Or it can be process-oriented, [affecting] all of the things that are happening in the supply chain.
Q: What has been the impact on wages?
Hicks: Wages have not changed a great deal. What I was referring to was the change in the composition of workers. Go back to 2000, and split manufacturing workers into three types — low-, medium- and high-skilled. Then use wages: low-skilled is under $15 an hour; medium-skilled is $15 to $20 an hour, and high-skilled is over $20. If you look at the number of people in those occupations [compared with] 2000, what you see is that the low-skilled occupations have 40-percent fewer employees. They got devastated by the Great Recession, by permanent normalized trade relations with China, by Nafta, whatever you want: those [jobs] are gone in big numbers. If you look at middle-skilled jobs, they’re down about 14 percent. But for jobs that were high-skilled in 2000, they’re up about 375 percent. They went from only a quarter-million to over a million and half jobs.
That change is exactly what the economic models that talk about trade and automation would suggest, because the U.S. is heavily endowed with high-skilled, high-wage workers. They have high school diplomas underneath their belt (although by 2007, half of all manufacturing workers had actually been to college), so there’s a comparative advantage in the very high-skilled, highly productive jobs, and a comparative disadvantage in assembly operations, low-value-added agriculture, and the like. That’s where the job losses have been clustered. It’s not just manufacturing jobs that have been lost; it’s a big transfer. I would argue that most of those high-wage manufacturing workers were not employed in manufacturing at the start of 2000. We didn’t see a big retooling of workers. What we saw was people leaving college and going into manufacturing, while the high school-related jobs just went away.
Q: I’m hearing that we’re experiencing an actual shortage of highly skilled workers in some manufacturing sectors. Is that true?
Hicks: I think so. With the careful caveat that economists like to say that shortages only occur when government gets involved — it’s really a mismatch between what workers are willing to work, and what businesses are willing to pay. We’ve documented that in a couple of studies on automation. Particularly in the higher-level jobs, there seems to be some shortage. I worry, however, that on occasion we might hear an anecdote from a business that there’s a worker shortage, when it simply may be that businesses are hoping to hire people at a wage that workers are not willing to work at. That’s not a shortage; it’s simply a business model that’s not consistent with the reality of labor markets.
Q: So regarding the 750,000 manufacturing jobs that you found the economy has added since the end of the recession, those are higher-end to a great degree?
Hicks: Yes. From 2009, the recession saw a dramatic drop in manufacturing employment. Even when employment didn’t drop, it effectively did — we didn’t make any cars in the summer of 2009, even though the big auto dealers continued to employ people, pay them to clean up the shop and that sort of thing. The recovery since then has been one that’s marked by a very different type of manufacturing worker. As I mentioned, 2007 was the year when half of all manufacturing workers had been to college. And so growth in educational attainment by manufacturing workers has been profound. We’ve seen a sharp increase in demand for higher-skilled workers, with a two-year degree or bachelor’s degree in an engineering or technology-related discipline. I think that trend is accelerating, although I’m not sure it’s always going to be that way.
Q: You say the U.S. manufacturing base is not in decline. That’s cold comfort to the people who have actually lost jobs. What do we say to them?
Hicks: The simple, raw fact is that manufacturing output in the U.S. is higher than it’s ever been, while manufacturing employment is one-third lower than it was when I was in high school in the late 1970s. So if you are a worker that has lost a job in manufacturing, it’s very difficult. The authors of the M.I.T. study that I mentioned have recently looked at the polarization of voting areas, and found a very strong correlation between manufacturing job losses exposed to trade and support for either Bernie Sanders or Donald Trump, who were both seen as rather extreme figures on trade issues. So it’s certainly carrying over into voting patterns and the way that public policy is being developed. If you live in Muncie, Ind., or anywhere else in the Midwest that has really been clobbered, it’s evident every day that you drive home from work.
Q: The inevitable conclusion of what you’re saying has to be that protectionism is not going to solve our nation’s problems, either economically or in the manufacturing sector.
Hicks: Right. Although I do think there is room for better trade deals. For example, when the President argues that he needs to have better deals with China or Mexico, bilateral trade deals are probably the route of the future. Multilateral trade deals sound really good. They’re great for State Department and Commerce Department negotiators. The problem is that the enforcement mechanisms for countries that violate those rules are so cumbersome and lengthy, that if you’re a business owner or investor, or work on a shop floor in an industry with a cheater, it could be a decade before you find any relief through the enforcement mechanism. Your job could be long gone, and your house foreclosed upon, with you looking for work somewhere else before the WTO weighs in. Bilateral trade agreements allow for a lot more rapid negotiation. The downside is that it’s harder to get some concessions in bilateral agreements, but I don’t think there’s going to be any great relief for this, in part because the problem that we have is being echoed in Mexico. Over the past several months, I’ve spoken with a number of people there. They’re very unhappy that manufacturing employment growth has been so slow in Mexico — it’s exactly the same share of overall employment than it was in 1994, and it’s concentrated in a couple of places. Productivity gains in manufacturing have been very rapid in Mexico, and in China as well. There are now growing rust-belt towns outside of Beijing, where productivity has eliminated.
Timely, incisive articles delivered directly to your inbox.